Click-Gaussian is a novel method for interactive segmentation of 3D Gaussians, enabling real-time manipulation of 3D scenes. It addresses the limitations of existing methods by learning distinguishable feature fields at two levels of granularity (coarse and fine) without extensive post-processing. The method uses Global Feature-guided Learning (GFL) to handle inconsistencies in 2D masks across different views, ensuring consistent and reliable feature learning. Click-Gaussian achieves segmentation accuracy and computational efficiency, with a processing time of 10 ms per click, significantly faster than previous methods. Extensive experiments on real-world datasets demonstrate its superior performance in both coarse and fine-level segmentation, making it suitable for various applications such as 3D scene editing and manipulation.Click-Gaussian is a novel method for interactive segmentation of 3D Gaussians, enabling real-time manipulation of 3D scenes. It addresses the limitations of existing methods by learning distinguishable feature fields at two levels of granularity (coarse and fine) without extensive post-processing. The method uses Global Feature-guided Learning (GFL) to handle inconsistencies in 2D masks across different views, ensuring consistent and reliable feature learning. Click-Gaussian achieves segmentation accuracy and computational efficiency, with a processing time of 10 ms per click, significantly faster than previous methods. Extensive experiments on real-world datasets demonstrate its superior performance in both coarse and fine-level segmentation, making it suitable for various applications such as 3D scene editing and manipulation.